The Asymmetric Bet: Over-Indexing on AI for Engineering and Enterprise

Alpine Investors & Portfolio Companies

Mo Battah

2025-08-20

Commandments for this Presentation

  1. Raise your hands! Your contributions are more important than my content, and we will finish my content anyway. I will auto-manage our time.
  2. Clarify and Correct: In this rapidly advancing world of AI, words can have multiple meanings. If I say something that sounds wrong, needs clarification, or is out of date, please correct me. Share your stories or complimentary information. This is a place where everyone has something to contribute; there are no know-it-alls, only learn-it-alls. Our diverse audience from private equity portfolio companies means diverse experiences, which is a good thing.

Session Guidelines & Opening Questions

  • This is a discussion—jump in via chat or hand raise.
  • How will the Software Development Life Cycle change when every engineer is 10x more productive but not 10x more knowledgeable?
  • How is QA evolving? Are developers spending more time reading and prompting than writing code?
  • Does AI change the ratio of developers to PMs or QA?

My Essential AI Tools

This section details the key tools that power my AI workflow.

Tool Deep Dive: Wispr Flow (Voice Dictation)

  • Primary Use: Voice-to-text dictation across platforms (phone, text messages, emails, AI interaction).
  • Volume: Dictating over 100,000 words per month, potentially nearing 200,000.
  • Workflow: Preferred method for interacting with AI. Dictate thoughts, AI structures them. Enables efficient message management (e.g., while walking).
  • AI Interaction: Voice-to-text input, text output from AI. Avoids voice-to-voice due to reduced AI content output.

Tool Deep Dive: Agentic Coding

  • Highest Abstraction: Claude Code (pioneer, model context protocol creator)
  • Secondary: Gemini CLI (for tasks not consuming Claude usage, due to hitting Claude limits easily)
  • Less Agency: Roo (not currently relevant for your software engineering level)

Tool Deep Dive: Context Storage and Management (Crucial)

  • Method: Massive GitHub repository storing context layers in Markdown files.
  • Why Markdown: AI prefers Markdown; aesthetically pleasing and integrates well with Quarto.
  • Types of Context: Career feelings, emotions, project goals, business insights, client/customer information, problem definitions.
  • Avoid: Storing context directly with foundation model providers.

Tool Deep Dive: Other Tools

  • Willow
  • Hypernote
  • (Your own custom solution - mention its complexity for cross-platform/feature sets)

Part 1: The New Mental Model

Over-Index on AI – It’s an Asymmetric Bet

  • $20/month tools can unlock outsized ROI.
  • Anecdotes:
    • Job search: ~$50 investment leading to a ~$40k salary increase.
    • $200 cloud credit enabling significant business value.

Augmentation, Not Abdication

  • AI augments judgment; it doesn’t replace it.
  • Technical due diligence: AI probes deeper, but humans make the final call.
  • If a task can be automated, find the next, more valuable task.

Part 2: The Practitioner’s Toolkit

The Ladder of Abstraction: From Prompting to Partnership

Level 1: Conversational AI

  • ChatGPT, Claude and similar tools as baseline assistants.

Level 2: Integrated Assistants (IDE Extensions)

  • Cursor, Roo, and similar VS Code extensions.
  • They read codebases, propose plans, and modify code with approval.

Level 3: Agentic Partners (CLI Tools)

  • Claude Code, Gemini CLI, Codex and others.
  • Operate within the terminal, execute scripts, and manage files.
  • Example: organizing a downloads folder or processing .eml email files.

Workflow: Knowledge Pipeline (Quarto & LaTeX)

  • Write in Markdown; render beautiful PDFs.
  • Reuse templates for SOWs, reports, and a flexible resume system.

Workflow: Information Edge

  • Export LinkedIn data and query it far beyond the UI.
  • “Deep research” prompts create rich briefings for meetings.

AI in Action: Automatic LinkedIn CRM

The Complete Workflow: - Input Flexibility: CSV export, screenshot, or simple copy-paste of LinkedIn connections - Deep AI Research: Claude Code investigates each person’s background, current role, and company context - Context Layer Integration: Cross-references against my personal context (career goals, business challenges, opportunities) - Strategic Reconciliation: Identifies potential collaborators, advisors, or business opportunities based on alignment - Automated Database: Builds enriched contact profiles with actionable insights and next steps

Key Insight: This transforms passive networking into active relationship strategy. Instead of hoping to remember who someone is months later, AI creates a living database that understands why each connection matters to your specific goals.

The Context Layer Dependency: This only works because I maintain detailed context files in GitHub—my career aspirations, current projects, industry focus areas. Without this foundation, AI can’t make meaningful connections.

Workflow: Personal Augment (Vocabulary Builder)

  • Inspired by: “Word Power Made Easy” by legendary grammarian Norman Lewis.
  • Functionality: Feed it your work calls, presentations, or documents. The AI helps you learn vocabulary at the very edge of your knowledge base, making your language more precise and impactful.
  • Advanced Application: An implementation based on “The Well-Spoken Thesaurus” by Tom Heehler. This analyzes your content, encouraging you to replace weak phrases and words with stronger, more meaningful, and active vocabulary.
  • The Power of AI: This demonstrates how AI can take a finite resource—a beloved book or a specific pedagogical approach—and extend its value infinitely. You can continue to benefit from an author’s profound insights and teachings, even if they are no longer with us, by applying their methods through AI to your own contemporary content. This transforms a static resource into a dynamic, personalized learning engine.

AI in Action: Folder Structure

AI Folder Structure

Part 3: The Broader Implications

AI as a Deflationary Force

  • Skydio drones inspecting power lines.
  • Gecko Robotics monitoring infrastructure.
  • Improved efficiency in the physical world lowers costs for everyone.

Beware of “Vibe Coding”

  • Productivity isn’t a substitute for foundational systems knowledge.
  • Example: choosing NoSQL when a relational database is required.

Conclusion & Discussion

  • Revisit the opening questions.
  • What shifts do you see in your teams and workflows?
  • Let’s talk.